It’s been ten years since Google coined the term “knowledge graph” and (broadly) laid out how its knowledge graph worked. Generally speaking, a knowledge graph represents semantics by describing the many entities (and their relationships) present within information.
By using a graph structure, e-commerce companies can showcase the items they want you to check out. For instance, say an individual is a die-hard fan of Tom Hanks. He or she may then like the movies “Forrest Gump” or “Sully” since Tom Hanks is in both of them. The individual in question could also be searching for something else outside motion pictures. It may be the book, “Sully: My Search for What Really Matters”, which inspired Clint Eastwood to adapt it for the big screen. If music is what our mysterious shopper is after, then Three Dog Night could be an option as the rock band is part of the Forrest Gump soundtrack.
To talk to us about ontology, knowledge graphs, and everything in between, we recently sat down with Katariina Kari, lead ontologist at Inter IKEA Systems B.V. and Scoutbee’s knowledge graph advisor. Originally from Finland and with a master’s in science, Katariina specializes in semantic web as well as guiding enterprises – of all kinds – to the ways in which digital business platforms can help them accelerate their transformation journey. Her knowledge graph credentials include the modeling of Zalando’s fashion knowledge graph, a common vocabulary to offer customers meaningful product views.
Scoutbee: Katariina, first things first. We really appreciate you taking the time to talk to us. Full disclosure, we had to google the term ontology. Please enlighten us: What does an ontologist do?
Katariina: An ontologist is a person who creates an ontology in a computer tech sense. In simpler terms, they describe to a computer the way humans see the world. When you think about machine learning for example, what it essentially does is: it looks at the data as evidence of human behavior and then it makes up its own mind, or its own inner model, about how it thinks the world operates.
Ontologists, on the other hand, say: “Hey computer, here are the implicit things that people know about the world and I am making them available explicitly to you so that you can make logical conclusions about the presented data.” In practice, it means that when you create a knowledge graph, the ontologist is responsible for its shape, its data structure (upon which the knowledge graph is built).
The one thing ontologists always need to keep an eye out for is that mistakes are not made when creating the structure. If an ontologist explains things to a computer on a 1:1 basis, there is no scaling up. So, ontologists create minimal but powerful facts to ensure that what is being described to computer systems (be it machine learning, or an Application Programming Interface – API, etc.) can be used to describe other things as well.
This way, it covers most of the areas that a knowledge graph should be expressing. Ontologists must ensure that what’s being described is something the system can actually use for deduction and logical conclusions so that the whole solution can be scaled up.
Scoutbee: Would you say then that ontologists bridge the gap between a user and the engineers creating the applications and systems, and navigate well between tech and other departments within companies?
Katariina: Absolutely. What engineers are realizing now is that ontologists are very useful within teams, so that they don’t need to talk to other departments or to stakeholders at all. The ontologists will do that, describing everything that has been shared in a data format so that the engineers only need to query what the ontologist has created and off they go.
Scoutbee: What’s the ideal skill set for the role?
Katariina: Usually, the profession requires a mix of computer programming in combination with areas such as sociology, aesthetics, philosophy, and communications. A background in linguistics, especially computational linguistics, is also very useful. Having said that, if you just happen to have a technical background but like social interaction, you will do just fine.
Scoutbee: Knowledge graphs seem to be very close to your heart. Tell us what they are useful for, and why they are important in the context of digital transformation?
Katariina: Knowledge graphs are really good at creating a common understanding of data in an enterprise. They’re great at breaking data silos: for example, when you have things stored in databases – for reasons such as legacy or performance – it often lacks context. You’ve got this sea of information lying around but may not always know what the data is for, why it has been stored, or even what kind of data it is and what it relates to. So, we tend to refer to knowledge graphs as the semantic layer that goes on top of your data: something lightweight that basically explains everything, and gives context to all the stored data.
They’re also very useful for what we call Explainable AI. For example, by infusing machine learning processes with knowledge graphs, e-commerce stores can recommend things to customers – and explain why they think their customers may like them. Overall, knowledge graphs improve machine learning performance by incorporating all the related data that an application needs. A practical example of this: Google uses knowledge graphs to figure out whether you meant Jaguar (the car) or the largest of South America’s big cats when you first typed “jaguar” into the search engine.
Scoutbee: I love that example. Let’s try bringing it closer to our customers’ reality. Say I am a procurement manager working for company X that has tons of supplier data stored in its ERP system. How are knowledge graphs going to help me?
Katariina: Anyone managing a large supplier database will agree that suppliers have their own creative way of expressing their information. So, you may end up having dozens of tabs displaying loads of information shared by thousands of suppliers.
With knowledge graphs, you have a common language to make sense of your suppliers: once you map the supplier output (the way they describe themselves, industry-related terms or disclose their info) for the first time, any future data provided by them will be comparable: you make them all speak the same language.
Scoutbee: Does that mean that every single one of our existing and future customers (from any given industry) can benefit from Scoutbee’s knowledge graph in the context of acting and decision making?
Katariina: Absolutely. It’s like standards: it benefits a wider range of customers and sectors. It does, however, depend on how you make it available to your customers. Broadly speaking, Scoutbee’s knowledge graph enables all suppliers to speak the same language and you, as a procurement leader, don’t have to be “fluent” in all the languages used by your suppliers but rather have a comparable view of them.
Scoutbee: Our customers will be pleased as this will make their processes even simpler. Now, you are joining us following the release of SCOUTBEE Supplier Intelligence, a SaaS-solution for procurement that uses AI and Scoutbee’s knowledge graph to map data from multiple areas, including – but not limited to – Scoutbee’s 10 million supplier profiles, extensive internet search, and third-party supplier data. Why have you decided to join Scoutbee?
Katariina: I’m sorry to say that data engineers still cannot solve all the problems and that data-driven business models are here to stay – which is why I’m so excited to be joining Scoutbee. Given that Scoutbee operates in a space that tends to be quite traditional, with markets still dominated by this old data engineering mindset, I’m amused to see Scoutbee investing in knowledge graphs.
I’d love to play a small part in the journey. Since I have walked the knowledge graph path a few times, I believe I’ve gathered enough experience to know what works and what doesn’t. I’m hoping to accelerate Scoutbee’s natural learning curve. I am also very knowledgeable of standards in general so that’s another asset I bring to the table.
Scoutbee: Last but not least. A Deloitte study has shown that women’s share in the overall global tech workforce has increased by 6.9% between 2019 to 2022, while their share in technical roles has grown by 11.7%. Have women in tech finally cracked the glass ceiling? What advice would you give to women wanting to go into the industry?
Katariina: Women in tech are stepping up. There are more and more females going into the profession. We need to keep encouraging them to do so. But, before that, tech needs to change. It needs to be more inclusive, more humane, a profession that can accept many truths, one in which you don’t feel you always have to be sure of everything. Good tech is not created by being sure. The tech field is already changing but it needs to change faster to become a well-balanced environment to work in.
My advice to female professionals in tech is: always ask questions, dare to be different, explore your differences, don’t just fit in. In fact, I’d give the same advice to anyone, regardless of their gender identity. Tech would benefit enormously from having more varied perspectives.
Ask me anything
You’re happiest when? When I’m pottering around the house, doing leisure-related things.
Any favorite lines from a movie? Pretty much any line in Bridget Jones’s Diary is hilarious. I love the sentence: “Great. I am wearing a carpet.” I relate to it when I’m wearing something that I don’t feel comfortable in.
Favorite travel spot? My parents’ summer house in Ingå Archipelago, South Finland.
How do you define success? Feeling happy and comfortable doing something.
If you could do another job for just one day, what would it be? I would be a sourdough baker in a cute village. I love sourdough baking – and cute villages.
If you could meet anyone in the world, dead or alive, who would it be? Any of the known religious philosophers.
If you had to eat one meal every day for the rest of your life, what would it be? Mozzarella tomato cucumber salad with herbs (that’s basically what I did in my 20s). Or a nice Japanese-style chicken rice.
What are your three most overused words/phrases? “Let’s cross that bridge when we come to it.” “Don’t boil the ocean.” “Where is the pull request?”